This tutorial walks through the steps for reproducing Bundle Analytics 1 results on Parkinson’s Progression Markers Initiative (PPMI) 2 data derivatives. Bundle Analytics is a framework for comparing bundle profiles and shapes of different groups. In this example, we will be comparing healthy controls and patients with parkinson’s disease. We will be using PPMI data derivatives generated using DIPY 3.
First we need to download streamline atlas 4 with 30 white matter bundles in MNI space from
For this tutorial we will be using a test sample of DIPY Processed Parkinson’s Progression Markers Initiative (PPMI) Data Derivatives. It can be downloaded from the link below
If you prefer to run experiments on the complete dataset to reproduce the paper 1 results please see the “Reproducing results on larger dataset” section at end of the page for more information.
There are two parts of Bundle Analytics group comparison framework, bundle profile analysis and bundle shape similarity analysis.
For generating bundle profile data (saved as .h5 files): You must have downloaded bundles folder of 30 atlas bundles and subjects folder with PPMI data derivatives.
Following workflows require specific input directory structure but don’t worry
as data you downloaded is already in the required format. We will be using
folder you downloaded from streamline atlas link and
downloaded from test data link.
Make sure all the output folders are empty and do not get overridden.
out_dir folder (eg: bundle_profiles):
Run the following workflow:
dipy_buan_profiles bundles/ subjects_small/ --out_dir "bundle_profiles"
For running Linear Mixed Models (LMM) on generated .h5 files from the previous step:
out_dir folder (eg: lmm_plots):
And run the following workflow:
dipy_buan_lmm "bundle_profiles/*" --out_dir "lmm_plots"
This workflow will generate 30 bundles group comparison plots per anatomical measures. Plots will look like the following example:
out_dir folder (eg: sm_plots):
Run the following workflow:
dipy_buan_shapes subjects_small/ --out_dir "sm_plots"
This workflow will generate 30 bundles shape similarity plots. Shape similarity score ranges between 0-1, where 1 being highest similarity and 0 being lowest. Plots will look like the following example:
Complete dataset of DIPY Processed Parkinson’s Progression Markers Initiative (PPMI) Data Derivatives can be downloaded from the link below:
Please note this is a large data file and might take some time to run. If you only want to test the workflows use the test sample data.
All steps will be the same as mentioned above except this time the data donwloaded
will have different folder name
subjects instead of
For more information about each command line, you can go to https://github.com/dipy/dipy/blob/master/dipy/workflows/stats.py
If you are using any of these commands do cite the relevant papers.
Chandio, B.Q., Risacher, S.L., Pestilli, F., Bullock, D., Yeh, FC., Koudoro, S., Rokem, A., Harezlack, J., and Garyfallidis, E. Bundle analytics, a computational framework for investigating the shapes and profiles of brain pathways across populations. Sci Rep 10, 17149 (2020)
Marek, Kenneth and Jennings, Danna and Lasch, Shirley and Siderowf, Andrew and Tanner, Caroline and Simuni, Tanya and Coffey, Chris and Kieburtz, Karl and Flagg, Emily and Chowdhury, Sohini and others. The parkinson progression marker initiative (PPMI). Progress in neurobiology, 2011.
Garyfallidis, E., M. Brett, B. Amirbekian, A. Rokem, S. Van Der Walt, M. Descoteaux, and I. Nimmo-Smith. “DIPY, a library for the analysis of diffusion MRI data”. Frontiers in Neuroinformatics, 1-18, 2014.
Yeh F.C., Panesar S., Fernandes D., Meola A., Yoshino M., Fernandez-Miranda J.C., Vettel J.M., Verstynen T. Population-averaged atlas of the macroscale human structural connectome and its network topology. Neuroimage, 2018.